EP 3882824 A1 20210922 - ADAPTIVE QUANTIZATION METHOD AND APPARATUS, DEVICE AND MEDIUM
Title (en)
ADAPTIVE QUANTIZATION METHOD AND APPARATUS, DEVICE AND MEDIUM
Title (de)
VERFAHREN UND VORRICHTUNG ZUR ADAPTIVEN QUANTISIERUNG, VORRICHTUNG UND MEDIUM
Title (fr)
PROCÉDÉ ET APPAREIL DE QUANTIFICATION ADAPTATIVE, DISPOSITIF ET SUPPORT
Publication
Application
Priority
- CN 201811358824 A 20181115
- CN 2019106084 W 20190917
Abstract (en)
Provided are an adaptive quantization method and apparatus, a device and medium. The method comprises: respectively performing a first quantization processing on a plurality of original input tensors to obtain an input tensor in a fixed-point number form, and calculating a quantization offset of the input tensor in the fixed-point number form (S102); calculating a comprehensive quantization offset corresponding to the plurality of original input tensors, and an adaptive quantization coefficient (S104); according to the adaptive quantization coefficient and the comprehensive quantization offset, performing a second quantization process on the input tensor in the fixed-point number form and the quantization offset to obtain a quantization result (S106). The method is helpful to improve the quantization accuracy, improve the performance of the convolutional neural network, and reduce the hardware power consumption and design difficulty.
IPC 8 full level
G06N 3/08 (2006.01)
CPC (source: CN EP KR US)
G06F 7/49947 (2013.01 - US); G06F 7/57 (2013.01 - US); G06F 17/15 (2013.01 - CN KR); G06F 17/16 (2013.01 - EP); G06N 3/045 (2023.01 - CN EP KR); G06N 3/0495 (2023.01 - KR); G06N 3/063 (2013.01 - EP); G06N 3/08 (2013.01 - KR); G06N 3/082 (2013.01 - CN); G06F 17/15 (2013.01 - US); G06N 3/08 (2013.01 - US)
Designated contracting state (EPC)
AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR
Designated extension state (EPC)
BA ME
DOCDB simple family (publication)
EP 3882824 A1 20210922; EP 3882824 A4 20220831; CN 111191783 A 20200522; CN 111191783 B 20240405; JP 2022507704 A 20220118; JP 7231731 B2 20230301; KR 20210093952 A 20210728; US 2022091821 A1 20220324; WO 2020098368 A1 20200522
DOCDB simple family (application)
EP 19883762 A 20190917; CN 201811358824 A 20181115; CN 2019106084 W 20190917; JP 2021527128 A 20190917; KR 20217018288 A 20190917; US 201917294432 A 20190917